SCACP Counters

Sai Krishna Manian
sai-krishna.manian@hp.com

Introduction

Systems Communications Architecture Control Program (SCACP) is the primary utility to monitor and manage a node in an OpenVMS cluster. SCACP provides a window to peep into the current state of cluster communication between the nodes through the varied counters it makes available to the administrator. These counters are the first piece of evidence for any issues with the cluster communication. The document describes the various counters in OpenVMS clusters seen through the SCACP commands –‘SHOW CHANNEL’, ‘SHOW VC’, and ‘SHOW LAN’. The SCACP counters for all these commands are explained in the sections below.

SHOW CHANNEL Command

This section describes the various SCACP counters for SHOW CHANNEL command. It includes the following sections:

• Channel Summary
• Channel Miscellaneous
• Channel Equivalent Channel Set
• Channel Counters and Errors
• Channel Errors

Channel Summary

Channel state

The channel can be in one of the three states,

• Open
• Path, or
• Closed.

When the channel handshake is done, the channel is at Path state and, if useable will transition to Open state. It is at Close state, when the channel is unusable or the remote node is down.

Total errors

The summation of all the errors listed in Channel Errors for this channel and the number of retransmits and transmit failures. For more information on channel errors see “Channel Errors”.

ECS state

It denotes the Equivalent Channel Set state of the channel. The channel is an ECS member, if the state is ‘Y’. The membership is based on three criteria. For more information see “Channel Equivalent Channel Set”.

Priority: current

Current priority used to evaluate the channel for ECS membership. This is the sum of management priority values assigned to the LAN device. The priority can take values between -128 to +127. Refer [1]

Priority: management

Dynamic management assigned priority.

Hops

The number of network interconnection devices in this channel’s path to the remote node. By default, it is set to 2, if the value is not set by dynamic management. This field is not significant, currently.

Buffer size

The maximum size of data that can be contained in a packet sent over the channel. It is taken as the smallest of the following values, local LAN device buffer sizes, Remote LAN device buffer sizes, local and remote NISCS_MAX_PKTSZ system (SYSGEN) parameter values, and the largest packet size the intervening network can deliver, which is determined by the NISCA channel packet size probing algorithm. For IP channels, NISCS_UDP_PKTSZ (SYSGEN) parameter is used to control buffer size instead of the above mentioned parameters.

The size probe handshake is initiated by calculating the number of bytes to add to a standard Verify (Verf) packet to make it large enough to represent the desired trial packet size. Receipt of the Verify Ack (Vack) packet within the handshake timeout interval indicates that the channel will support packets of the trial size and the max buffer size is updated accordingly. The size probe timeout interval is significantly increased once the channel’s packet size limit is found.

With LAN channels, the common buffer size would be 1426 bytes (without jumbo frames). And with IP channels the buffer size is 1394 bytes.

Delay

Running average of the measured roundtrip time, in microseconds, for packets sent over the channel. The channel’s average Round-Trip Time (RTT) measurements can be taken either at the Transport (TR) layer or at the Channel Control (CC) Layer (in the absence of Virtual Circuit traffic).

At the TR Layer, the sender keeps the transmit time and the channel selected in its local transmit cache. The receiver records the received packet’s sequence number as last accepted packet. The receiver while sending includes the last accepted packet sequence number in the header. The sender would calculate the RTT or delay as the difference between the current time and the transmit time recorded in its cache.

At the CC layer, certain pairs of CC packets (CCStart-Verf and Verf-Vack) will yield accurate delay measurements.

The delay thus measured is combined with the current average RTT to obtain a weighted average. The measurement at the CC layer comes into picture in the absence of VC traffic over the particular channel. A delay probe timer initiates RTT calculation in the CC layer at regular intervals, but it checks if RTT was calculated at the TR layer since the last timer expiry; if yes, it just skips RTT calculation and resets the delay probe timer to the next interval.

The value is same as the Average RTTime shown in the next section (Equivalent Channel Set).

Load class

The data rate for this channel, initialized from local and remote LAN device speeds. It is set to the minimum of the local and the remote LAN device speeds.

Total packets

Total number of packets sent, including Channel Control (CC) packets, and received over this channel.

Ch open time

The last time this channel had a verified useable path to the remote node.

Ch close time

The last time this channel was closed.

Channel Miscellaneous

Buffer size: current

Current upper bound on buffer size for this channel.

Buffer size: negotiated

Negotiated maximum common buffer size. The lesser of local and remote BUS limits.

Buffer size: management

Dynamic management assigned max block data field size. Only the lesser of this (when <>0) or negotiated value is used. This facility is not available to users as of today.

Buffer size: local

The channel’s local buffer size.

Buffer size: remote

Remote system's buffer size (largest possible buffer size).

Buffer size: change

The number of times the buffer size was changed for the channel. This field is reported as ‘Topology change’ in the ‘Channel Errors’ section in OpenVMS 7.3-2 and before.

Seq number: local

Sequence number of the local channel. The channel sequence number starts with 1 and is incremented every time the channel is closed, so that the next time this channel is used the sequence number will be different. This will prevent most cases of old TR messages stored in the network from being received and processed.

Seq number: remote

The sequence number of the remote channel.

Channel Equivalent Channel Set

ECS State

A Channel that is Open is an ECS member, if the flag is set to ‘Y’. If the Open Channel is not in ECS, the flag is set to ‘N’. The state bit is left as blank (“ “) for a channel that is currently Closed. The ECS membership is based on three criteria,

• Loses: The channel is Tight(T), if the packet loss history is acceptable and Lossy(L), if the recent packet loss history makes it unusable.
• Capacity: The priority and buffer size of the channel are used to classify whether the given channel is Peer, Inferior, Superior or it is Unevaluated.
  • Peer (P) – Priority and buffer size both match the highest corresponding values of the set of Tight channels, entitling the channel to be an ECS member.
  • Inferior (I) – Priority or buffer size does not match the corresponding values of the set of Tight channels.
  • Superior (S) – Priority or buffer size is better than those of the corresponding values of the ECS member channels. This is a short lived transient state because it exists only while the ECS membership criteria are being re-evaluated.
  • Unevaluated (U) – Priority or buffer size, or both have not been evaluated against the ECS criteria, usually because the channel is Lossy.
• Speed: If the channel’s delay is among the best when compared to other available channels that are Tight and Peer, F (fast) is set for this field; else, it is set to S (slow).

A channel has to be Tight, Peer and Fast, to be in ECS.

ECS losses

Value representing the channel’s recent packet loss history. A decision on the channel’s state is based on this value. This value is increased with every retransmission or transmission failure and diminished with each Hello packet received and also with each successful transmission over the channel.

The losses value is used in conjunction with two thresholds, lossPromote and lossDemote, to decide on the channel’s quality. The default is zero and a channel can’t be in ECS if the losses value is more than 128.

ECS transitions

The number of times this channel went in or out of the ECS. ECS transitions happen when there is a state change in the channel.

Average RTT

Average measured round-trip time for packets sent over this channel. This is extensively used in ECS calculations. The average RTT value must be less than the speed demote value for the VC, only then the channel’s ECS Speed is set to Fast(F).

For example, if the VC’s speed demote value for a node is 90 x-units and one of the channels to that node has an Average RTT value of 100 x-units. This channel will not be in ECS, as the Speed category will get set to Slow (S).

Remote ring size

Size, in buffers of the sender’s datalink receive ring for this channel. This indicates the maximum number of packets that can be buffered in the datalink before being accepted by the Data Exchange(DX) layer, and is used by the local CC and TR layers in the calculation of its maximum transmit window size. The initial default value assumed is 8.

Remote receive cache

Size, in number of buffers, of the sender’s transport layer receive cache (a.k.a re-sequencing cache). The number of out-of-order packets that can be buffered by the sender on receive, and is used by the local CC and TR layers in the calculation of its maximum transmit window size. The initial default value is 32. On VMS versions prior to 8.3 it would be 31. The automatically calculated values for this field could be 32, 64 or 128. The value can be set to a maximum of 1024 using MC SCACP. See MCR SCACP HELP for additional information.

Time in open state

Total time that this channel was in Open state.

ECS member time

Total time that this channel was in the ECS.

Channel Counters and Errors

Xmt messages

Number of messages, including control messages, sent on this channel. The value is longword bound and it would wrap after reaching (2^32)-1.

Xmt bytes

Number of bytes transmitted on this channel, including the control messages. The value is longword bound.

Receive messages

Number of messages, including control messages, received by the channel. The value is longword bound.

Receive bytes

Number of bytes received by the channel. The value is longword bound.

Xmt:rexmit

The number of successful messages per re-transmission. Higher the value (towards Xmt messages) better is the channel. The maximum it can reach is the total number of messages transmitted on this channel. If there are no retransmit penalties, the field will be shown as “infinite”.

Rexmit errors

The number of retransmits over the channel. Retransmits happen when the retransmission timer for a packet expires. With every retry the channel is penalized. This can affect the channel’s state and thus its ECS membership.

TransmitFail Penalties

The number of transmit failures over the channel. These are at the data exchange layer. These errors are also accounted for channel’s lossiness and hence a change in channel’s state.

Receive errors

The total number of receive errors. The summation of all the receive errors seen under “Channel Errors”. For more information, see “Channel Errors”.

Other errors

The number of errors, other than the receive errors as seen in the “Channel Errors”. For more information, see “Channel Errors”.

Channel Errors

Receive: bad auth

Number of CC (Channel Control) packets received with a bad authorization field. The CC layer authenticates the node attempting to form the channel during channel initialization. If the authentication fails, this value is incremented.

Receive: bad ECO

Number of CC packets received with an incompatible NISCA protocol ECO revision field value. The ECO field in the channel structure specifies the Implementation revision level of the port code. It is set to zero whenever either the major or minor version field values are updated.

Receive: bad MCA

The number of bad multicast CC packets received. The received multicast message code can be a Hello, a Bye or a Solicit service. If the multicast message code is different, it is flagged as an invalid multicast message.

Receive: short

Indicates the number of CC packets received that were too short. The size of the CC message, excluding the header must be at a minimum value; this value changes with the version of NI-SCA implementation. If the size is less, it is flagged as a short message and is not processed.

Receive: incompat

Received CC packets that were incompatible with existing channels for this virtual circuit. The incompatibility could be due to mismatch in protocol numbers or the node names between the peer nodes.

Receive: old chan

Received packets from an old instance of a channel. Every time a channel is closed, it has a sequence number field which is incremented. This is to prevent any old TR messages which are floating around in the network from being received and processed.

Timeouts: handshake

Number of listen time-outs over the channel. No Hello message received within the listen time-out interval which is 8 seconds by default. In the absence of a Hello message within the timeout interval, the channel is closed. The Listen Timeout (ListenTMO) interval determines the reaction time in determining node or link failures, and hence, the cluster’s ability to initiate state transitions to recover from failures, in time. The Hello Timeout (HelloTMO) interval defines the frequency of the Hello messages.

The ratio of ListenTMO to HelloTMO determines the number of Hello messages sent during the ListenTMO interval. The recommended setting is:

HelloTMO * 2 <= 0.75 * ListenTMO

This allocates one quarter of the ListenTMO interval to network delay and insures that at least 2 Hello messages are lost before a listen timeout can occur.

By default, the Hello interval and listen timeout interval are set to 3s and 8s, respectively. They can be adjusted by using the dynamic SYSGEN parameter, PE4.

Bits 0-7 of PE4 are used for Listen timeout interval, with unit in seconds and bits 8-15 are used for Hello interval with unit in 0.1s (or 100ms).

If PE4 is set to 0x0703, the listen interval becomes 3s and the Hello interval will be set to 0.7s (or 700ms).

No MSCP server

The number of times a solicit service request was received without the MSCP server being loaded.

Disk not served

Number of times request was made for a non-served disk.

Path restart

The number of channel restarts. Channel is restarted on receipt of a CC handshake message indicating the channel has been closed remotely and the remote node is restarting the channel formation handshake. One such case is, receiving a CCSTART message when the channel is in Open state. This will initiate a path restart.

SHOW VC Command

This section describes the various SCACP counters for SHOW VC command. It includes the following sections:

• VC Summary
• VC Equivalent Channel Set Membership Criteria
• VC Sequence Numbers and Miscellaneous
• VC Counters
• VC Errors
• VC Summary

VC Summary

State

The current internal state of the virtual circuit can be Open, Path or Closed. If there is at least one channel open, then the VC is set to Open state. If there is at least one channel that has been established, but the VC has not yet transitioned, then the VC is set to Path state. If there are no usable channels available, then the VC is set to Close state.

Total errors

Number of times the virtual circuit has been closed or has had other errors, since boot.

Xmt:TMO

Ratio of total numbers of transmitted to retransmitted messages. The maximum it can reach is the total number of messages transmitted on this VC. If there are no retransmits, the field will be shown as “infinite”.

Channels open

Number of currently open channels available to the virtual circuit.

Channels ECS

Number of ECS channels currently in use by the LAN virtual circuit.

ECS priority

The minimum priority a channel must have in order to be included in the Equivalent Channel Set (ECS).

Max packet size

Maximum data buffer size in use by this LAN virtual circuit.

Rexmt TMO

Retransmission timeout, in microseconds. The length of time the virtual circuit is currently using to wait for an acknowledgment of the receipt of a packet before retransmitting that packet.

To obtain more accurate values for the timeout, the protocol measures the delay from the submission of a packet to the datalink to the reception of its Ack. These measurements, averaged over time, plus a multiple of the delay’s variance are used to estimate the largest reasonable value.

Xmt window: cur

Current value of the transmit window (or pipe quota). The number of packets sent before stopping to wait for an acknowledgment. After a re-transmission timeout (a packet loss is taken as a resultant of congestion), the transmit window is reset to a minimum to decrease congestion; it is allowed to increase as acknowledgments are received.

The window is held at the minimum size till a successful transmission. Then the window is gradually opened in a two—phase fashion. In the first phase (slow start), the window is opened relatively quickly, at a rate proportional to the rate at which the Acks are received. The window continues to grow till it reaches a safe size, which is calculated as a fraction of the window size at which the congestion was detected. With this the slow start phase is complete and starts the congestion avoidance phase. The window is not increased aggressively in this phase. The initial default value is 1 and it can scale till the max, which is covered in the next heading.

Xmt window: max

Maximum transmit window size currently allowed for the virtual circuit. The maximum sensible transmit window is determined by assuming that the transmitter has exclusive access to the network resources between the sender and receiver. To determine the maximum window size, it is necessary to consider buffering at:

The destination’s TR layer, the destination’s datalink layer, and any intermediate network device (this is provided through management input).

The first two items are exchanged via CC packets. The initial default value is 8.

Xmt window: mgt

The management specified upper bound on the transmit window size set using MCR SCACP. The value set by the command will override the automatically calculated transmit window size. See MCR SCACP HELP for additional information.

Xmt options

There are two transmit options currently available, packet checksumming (CKSM) and compression (CMPR). They are to be enabled using MC SCACP SET VC [/COMPRESSION][/CHECKSUMMING]. Please refer MC SCACP HELP for additional information.

Total pkts (S+R)

Total number of packets sent and received by the VC. The value is longword bound.

VC open time

The most recent time the VC was opened.

VC close time

The most recent time the VC was closed.

VC Equivalent Channel Set Membership Criteria

Number epochs

Number of times all ECS membership criteria was recalculated. Epochs will establish ECS parameters afresh for the channels. It will re-compute the priority, hops, and buffer size for the ECS. The events which cause the recalculation of the acceptable ranges of the capacity criteria (a new epoch) are:

• VC is closed and a tight channel becomes available
• A tight, superior channel becomes available
• A peer channel becomes closed
• When ECS becomes empty.

Number new ECS

Indicates the number of times the ECS was formed by selecting all active channels based on previously established ECS membership criteria.

Buffer size: VC

Maximum data buffer size for this VC.

Buffer size: ECS

Indicates the current ECS buffer size. This is set to the maximum of the buffer sizes of the channels. It defines the maximum size of the data that fits in a LAN frame.

ECS priority

Indicates the current highest priority of the channels in ECS.

ECS Hops

Minimum number of Hops for a channel in ECS. Set to the minimum of the hop values of the channels, which are Tight and Bus online. Hops define the number of switches or bridges in the channel's network path to the remote LAN device. This value is not of any significance, currently.

Load class

Set to the minimum of the ‘load class (in Mb/s) of the channels in ECS’ times the ‘minimum of the number of local and remote data-links’.

For example - If the minimum load class of a channel from Node X to Node Y is 100 and node X has 2 data-links and node Y has 3. Then the Load class here would be 100*2.

Current minimum delay

Current minimum average RTT across the current ECS channel members to the particular node. The value is used to establish the two speed thresholds, speedPromote & speedDemote, for the VC. The two threshold values are used to provide hysteresis in the classification of a channel as fast or slow. A fraction of the current minimum RTT is used as a window around the maximum acceptable round trip time to achieve the desired degree of hysteresis.

ECS speed threshold: faster

Current threshold for reclassifying a channel as FASTER than the current set of ECS channels.

ECS speed threshold: promote

Current threshold for reclassifying a SLOW channel to FAST. Slow channels whose round trip delays fall below this value are reclassified as “fast” and may be eligible for membership in the ECS.

ECS speed threshold: demote

Current threshold for reclassifying a FAST channel to SLOW. ECS member channels whose round trip delays exceed this value are removed from ECS.

ECS Speed threshold: management

A management-specified lower limit on the maximum delay (in microseconds) an ECS member channel can have. This field can be modified by MCR SCACP using SET VC with the ECS_MAX_DELAY qualifier. See MCR SCACP HELP for additional information.

# LAN device: local

Number of unique local LAN devices across the ECS.

# LAN devices: remote

Number of unique remote LAN devices across the ECS.

VC round trip time

Average round-trip time, in microseconds, for a packet to be sent and acknowledged. VC round trip time values are dependent on the delayed ACK or the ACKholdoff delay (which is up to 100ms). The VC round trip time is also dependent on the network traffic. The ACKholdoff delay can be changed using PE4 SYSGEN parameter. Bits24-31determines the ACKholdoff delay value in units of 10ms. The VC round trip time can be seen as ACK delay time and it is used in the calculation of the retransmit timeout for the VC. Whenever a new ACK delay is measured, it is compared with the current estimate of the ACK delay. The difference is a measure of the error in the delay estimate (delayError). A fraction of this delayError is used as a correction to update the current estimate of average ACK delay.

VC round trip deviation

Average deviation, in microseconds, of the round-trip time. The average of the absolute value of the error in the mean (mean deviation) is used as an estimation of the delay’s variance. The difference in the delay error (mentioned above) and the current estimate of the deviation is taken as deviation error. A fraction of this error is used as a correction to update the current estimate of the average round trip deviation. This value is also used in the calculation of the retransmission timeout.

VC Sequence Numbers and Miscellaneous

Next seq to send (NSU)

The sequence number to be used for the next message to be transmitted. Part of VC’s transmit window state. The sequence number is a 32-bit value. It starts with one when the VC is opened and incremented by one for each VC packet created. When the value reaches (2^32)-1, the number wraps to 0 and then the sequence is repeated indefinitely.

Lowest Ack Received (LAR)

The sequence number of the last ack received. It indicates the current trailing end of the transmit window.

Highest Ack acceptable (HAA)

The sequence number of the most recent message sent over the VC. This defines the highest the ACK acceptable.

Last accepted

The sequence number of the latest message accepted that is within the cache range. It is used in the channel delay calculation as seen before.

High seq received in order (HSR)

The sequence number of the currently received highest in-order packet at the PPC layer. This is to handle out-of-order delivery of messages. A packet is stored in the TR re-sequencing cache only if the difference between the received packet’s sequence number and the HSR value is less than the maximum size cache, else the packet is dropped.

Total receive and cached

The number of messages cached. This is same as "Cached" mentioned in “VC Counters”. It is just an incremental value which tracks the total number of packets cached by the VC, since creation.

Receive window max

Indicates the number of messages that can be cached by the transport layer. In other words, limits the number of out-of-order messages that can be received at a particular instance by the VC. By default its value is 32. When a change in load class is detected, this would be set to 32, 64 and 128 for 100Mbps, 1Gbps and 10Gbps channels, respectively. This value is passed in CC packets to the remote node to determine its transmit window size to this node.

Receive window mgt

Indicates management setting of the receive window size, if any this over-rides the default setting. The value can be set using MCR SCACP SET VC/WINDOW=RECEIVE=n.

Receive window rmt

The maximum receive cache size of the remote node. This value is used to determine the transmit window for this node to the remote node.

CmdQ len

Reflects the number of Port commands seen in the queue currently. The "MaxCmdqLen" in “VC counters” is set to the maximum value reached by this field.

Rsvp threshold

Number of packets sent before requesting the remote node to immediately return an acknowledgment. This is done by setting the RSVP flag in the TR header for the message. The transmit window size is called pipe quota. The RSVP threshold is set to half of the current transmit window size to the remote node. The RSVP flag can also be set when the transmit window is full.

Min rmt dl buffers

Indicates the minimum size of the communicating node’s (remote) datalink receive ring buffer. This value, multiplied by the number of devices in ECS, is used to determine the transmit window size to that node. The minimum value would be 8, which is the default, and it can’t exceed 255.

VC Counters

Transmit: Messages

The number of transmissions completed on this VC. Like most other counters this is also longword bound.

Transmit: Acks

The number of ACKs sent over the particular VC.

Transmit: Retransmits

The number of retransmits over this VC (same as "SeqPkt Timeouts" in VC Errors). The VC is closed if a packet is retransmitted more than a predefined number of times, which is currently set to 30. The retransmission timer uses the exponential back-off mechanism, by which the retransmission timeout is increased by a power of 2 for every retransmit of the packet. A packet which is retransmitted for the ‘n’th time will have the retransmission timeout value set to 2^n times the original value. Retransmits are assumed to be the resultant of congestion.

Transmit: bytes

Cumulative size of all the messages transmitted over the VC, measured in bytes. The value is longword bound.

Maxcmdq winfull

Indicates the number of times the transmit quota (or, pipe quota) flag has been set for a particular VC. The flag is set either when the Pipe quota has reached its maximum or when a packet is to be retransmitted. This flag prevents the transmit window to grow.. The flag is cleared when an Ack is received. The maximum transmit window (or pipe quota) can be set using MCR SCACP SET VC / WINDOW=TRANSMIT=n.

Maxcmdq len

Reflects the maximum or highest number of Port commands seen in the queue so far.

Receive messages

The number of messages received over this VC. The value wraps around after reaching (2^32)-1.

Receive: Acks

The number of ACKs received for transmits over this VC.

Receive: Duplicates

Reflects the number of duplicate messages received. Duplicates are received if the sender has not seen an ACK before the retransmission timer expires.

Receive: Bytes

Cumulative size of all the messages received over the VC and it is a 32 bit field.

Receive Cached

The number of messages cached. A message is cached only if it is not the next expected message and it is within the maximum cache range. These are determined based on the sequence number. Same as seen in “VC Sequence Numbers and Miscellaneous”.

Receive MSG:DUP

Ratio of "Messages Received" and "Duplicates" received, rounded to highest integer. "Infinite", if there are no duplicates.

VC Errors

Receive short message

The number of short transport messages. The size of the TR message, excluding the header must be at a minimum value; this value changes with the version of NI-SCA implementation. If the size is less, It is flagged as a short message and is not processed.

Receive Illegal ack

Reflects the number of illegal ACKs, the sequence number of which is higher than the Highest ACK Acceptable (HAA), received.

Receive illegal seq pkt

The number of illegal sequence packets. It can be either a sequence message, received when VC is closed or an un-sequenced TR message, when the VC is open.

Receive checksum failures

The number of TR messages with invalid checksum. Network induced data corruption.

Receive Cache misses

This field counts the number of messages which are not cached. The message might not be cached based on the following conditions:

Sequence number is out of cache range. If the sequence number difference from that of the next expected packet is greater than the max cache for the particular VC, then the message is discarded.

Datalink driver needs this VCRP back immediately (local datalink).

Receive FreeQ empty

This reflects the number of times the MFREEQ (in CC/TR)/DFREEQ (in CC/TR) became empty. When a connection is established, a certain number of buffers are allocated to receive incoming Message or Datagram. When these pre-allocated buffers are exhausted, the PORT interrupts the Port driver to allocate buffers from the non-paged pool.

Transmit Seq Pkt Timeouts

The number of retransmissions over this VC. A retransmission happens when the sender of the packet has not received an ACK for the packet within the retransmission timer interval. A single packet is attempted to be retransmitted 30 times after which the Transport layer closes the VC. The retransmission timer is increased exponentially at every attempt for the packet.

Transmit NoXmtChan

The number of times channels were not available for transmitting.

VC Closures Seq Pkt Retry Exh

The number of times VC got closed due to retry exhaustion. The VC has a constraint in the number of retransmits. The value is decremented every time a retransmit happens and the VC is closed when it reaches zero.

VC closures no path

The number of times the VC got closed due to the absence of any channel in ECS.

VC Closures buffer size Decr

If the retransmit cache contains at least one buffer whose size is incompatible with the new ECS' buffer size, the VC is closed. This field indicates the number times that happened. This can happen after a new ECS order has been established. It will re-compute buffer size for the ECS. At times the buffer size can go down, in such case, if there is an entry in the retransmit cache with higher buffer size requirement, then the VC is closed.

VC Closures LAN Xmt TMO

Account for VC close due to transmit completion wait exhaustion. The sends happen asynchronously. This happens when the local datalink has not completed transmitting the packet even after ample number of transmit-wait delays. The VC is closed.

VC Closures Non Paged pool

Number of VC closures due to unavailability of non--paged pool. Increasing the non--paged pool might solve this issue.

SHOW LAN Command

This section describes the various SCACP counters for SHOW LAN command. It includes the following sections:

• Device Summary
• Device Counters and Errors
• Device Errors
• LAN Counters and Errors

Device Summary

Device name & Type

The name of the LAN or IP bus. The device type field is specific to the NIC used and this field is not applicable for IP bus.

Errors + Events

Total of all transmit and receive errors and error events over this bus. The datalink layer provides NI-SCA with one or more access points to physical transmission media. Each access point is called a bus. NI-SCA must request services of the datalink layer in terms of a particular bus.

Status

The device status can be a combination of the following:

Not in use by SCA: BUS has been stopped. It is done with MCR SCACP by using STOP LAN or STOP IP command. See MCR SCACP HELP for additional information.

Mgt priority

Dynamic priority assigned for the Bus. This can be modified by MCR SCACP using SET LAN_DEVICE or SET IP_INTERFACE with the PRIORITY qualifier. See MCR SCACP HELP for additional information.

Buffer size

Maximum data buffer size for this Bus.

Mgt max buf size

Dynamic management assigned maximum block data field size. The value can be set by using the following command, MCR SCACP SET LAN /BUFFER=

Line speed

The device's data rate in Mb/s. The field will be shown as “N/A” when the bus is not in use by SCA.

Total Pkts (S+R)

Total number of messages transmitted and received over this bus. The value is longword bound. For individual numbers see the next section (Device Counters and Errors).

Current LAN/IP Address

The current LAN/IP address of this Bus.

Device Counters and Errors

Transmit messages

Number of messages sent by this BUS, includes Multicast messages. The value is longword bound and it would wrap after reaching (2^32)-1.

Transmit bytes

Number of bytes sent by this BUS, includes Multicast messages. The maximum value of this field is (2^32)-1 and it wraps after that.

Transmit errors

Summation of generic transmit and HELLO errors (as seen in 'Device Errors').

Receive Messages

Total number of messages received by this BUS, includes Multicast messages. The boundary is similar to that of the ‘transmit messages’ field given in this section.

Receive bytes

Total number of bytes received by this BUS, includes Multicast messages. The limit is similar to that of the ‘transmit bytes’ field given in this section.

Receive errors

The sum of all receive errors seen on this BUS (as seen in 'Device Errors').

Port usable

The total number of PORT_USABLE events, that is, the data-link changes to Up/Online state.

Port down

The number of PORT_UNUSABLE events occurred, that is, the data-link changes to Down/Offline state.

Address change

The number of address change events in the LAN/IP device.

Restart failed

The number of restart failures on the LAN/IP device.

Last event time

Indicates the latest timestamp of any of the above events (marked with a '*').

Device Errors

Transmit generic errors

Count of local hardware transmit request that had problems or errors.

Transmit hello errors

Count of transmit errors during HELLO.

Last transmit error

Timestamp of the last transmit error.

Received with Bad SYSTEMID

Number of messages received with a wrong SCSSYSTEMID. For example, if a node tries to boot into a cluster with a SCSSYSTEMID which is already held by another node in the cluster, this counter is incremented.

Received Mcast TR message

Count of the Multicast messages directed to the Transport layer.

Received short CC msg

Number of short CC messages received.

Received short DX msg

Count of short DX messages received.

Received on Wrong port

Count of messages addressed to the wrong PORT, caused by either bad group code or SCSSYSTEMID /multicast address.

Received Discard port disable

Count of messages discarded because the PORT was disabled.

Allocation failures on CH

Count of channel allocation failures.

Allocation failures on VC

Count of virtual circuit allocation failures.

LAN Counters and Errors

Transmit packets

Total number of packets sent over this LAN.

Transmit bytes

Total number of bytes sent over this LAN.

Transmit errors

Total number of transmit errors over this LAN.

Receive packets

Total number of packets received over this LAN.

Receive bytes

Total number of bytes received by this LAN.

Receive errors

Total number of receive errors over this LAN.

Buffer unavailable User

Receive discards due to user buffer unavailability.

Buffer unavailable System

Receive discards due to system buffer unavailability.

Unrecognized individual

Receive discards due to unrecognized individual destination.

Unrecognized multicast

Receive discards due to unrecognized multicast destination.

Events

Total number of LAN events.

For more information

• “MCR SCACP HELP” is an important resource.
• Cluster Troubleshooting
• Troubleshooting the NISCA protocol